hahnjo wrote:
> I could allocate some time to debugging this, and here's what I got.
Thanks for trying all that!
> It looks like it happens as you suspected ("`D->hasExternalLexicalStorage()`
> returns `false` and we skip the call to `CompleteType`").
Ok, at least the understanding of the problem is progressing.
> I tried changing that condition to `getExternalSource() &&
> !D->getDefinition()`, but it doesn't help - I'm still getting the same
> assertion:
Too bad, that would have been too easy I guess... Now I wonder if we maybe need
to change the order of operations, ie first complete the redecl chain, then
notice that we can get the definition externally, and then actually go and do
that. Conceptually something like
```c++
// Complete the redecl chain (if necessary).
D = D->getMostRecentDecl();
if (D->hasExternalLexicalStorage() && !D->getDefinition())
getExternalSource()->CompleteType(const_cast<RecordDecl*>(D));
```
Not sure if that changes anything (at least it doesn't fail Clang tests when
applied to current `master` on its own), but as I said earlier "redecl chains
are a bit black magic for me". But if we are getting close to getting a
sharable reproducer, I will be able to have a look myself.
> Please let me know, if you have any more ideas I could try.
Only if you have more time to spare, the second part of point 4. above, ie
which of the commits is the culprit: Is it a single one and everything is fine
if you just exclude / revert that one?
https://github.com/llvm/llvm-project/pull/133057
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
